Designing dual linear power supply

[Vovk_Z]

It will work but there will be an unbalanced as per your arrangement .

No. It won't be unbalanced.

[Vovk_Z]

if I've made any terrible mistakes.

I don't see any critical mistakes. In general, it is ok.
An only possible problem: It was mentioned yet that 15 VAC is not enough to have a well-regulated 15 VDC. So you need at least 18 VAC secondaries.

[Vovk_Z]

The design goal was a power supply with low noise,

Low noise design is a bit another and deeper thing. It is a bit lazy for me to start one more topic here about how to do low noise regulated power supply (I'm sorry for that).

[bdunham7]

I will make this change.

I have also seen some power supplies with a 2watt or larger resistor with a capacitor across the rails. Should this be considered? Is there a rule of thumb for this?

What output current are you aiming for?  If you need 0.5A or less, your input caps are plenty big and the transformer will likely be fine.  If you need more current, you will have to be more detailed as to the transformer's specifications.  If the datasheet is showing 1uF tantalums, I would not follow that.  For your output, I think a 10uF electrolytic and a 0.1uF film cap would be good.  For the input, a low-ESR electrolytic over 1000uF plus a 0.1uF film cap will work as well.

I'm not sure what you mean about the resistor across the rails.  Perhaps this is to give the regulator a minimum load?  Can you explain this better or draw it?

[networkingdude]

It *looks* like a really interesting tradeoff. Add another rectifier, but gain the benefit of using identical positive regulators on both top/bottom, with more component choice.

A question, more than a comment. Is there a difference in the ripple phase, between this and the more conventional single rectifier approach? Here, the ripple would be 180 out of phase - which is desirable I think - for stuff like powering bipolar supply ops. eg. the common mode is cancelled.

The LM350 has no ideal negative counterpart (That I could locate anyway) so this appeared as the best solution that gives the most flexibility with lower noise I hope. Being out of phase and it helping is over my head at this point

[networkingdude]

I will make this change.

I have also seen some power supplies with a 2watt or larger resistor with a capacitor across the rails. Should this be considered? Is there a rule of thumb for this?

What output current are you aiming for?  If you need 0.5A or less, your input caps are plenty big and the transformer will likely be fine.  If you need more current, you will have to be more detailed as to the transformer's specifications.  If the datasheet is showing 1uF tantalums, I would not follow that.  For your output, I think a 10uF electrolytic and a 0.1uF film cap would be good.  For the input, a low-ESR electrolytic over 1000uF plus a 0.1uF film cap will work as well.

I'm not sure what you mean about the resistor across the rails.  Perhaps this is to give the regulator a minimum load?  Can you explain this better or draw it?

Like this

I'm aiming for 2A output x 2. I understand I need to increase transformer voltage to 18V if I want regulated 15v out.
The data-sheet said 0.1μF ceramic or 1uF tantalum on the input that's why I chose it. The 5600uF was close enough to 6000uF for rule of thumb of 2000uF per amp for ripple rejection.

EDIT: Transformer is 35VA 18V Talema 70074K I think is what I will use. I have not ordered parts or anything.

[iMo]

I built myself a 2x15V/50mA and 2x5V/100mA fixed voltage lab power supply on a single 10x10cm pcb. With two potted pcb transformers on it, each with 2 secondary windings, with 4 dil8 diode bridges in total. 2xLM317 for 5V and 2xLM723 for 15v (with 50mA foldback). All 4 sources are "galvanic" isolated, thus I have 4 voltages with 4 grounds. You may interconnect those sources as you wish (imagine 4 separate batteries with 15V, 15V, 5V, 5V). With a lab voltage source there is always the problem how to limit the current such you do not smoke your chips when experimenting. With bigger pcb I would use 4x 723, or similar, in order to provide a current limiting with foldback at each voltage source..

[networkingdude]

I built myself a 2x15V/50mA and 2x5V/100mA fixed voltage lab power supply on a single 10x10cm pcb. With two potted pcb transformers on it, each with 2 secondary windings, with 4 dil8 diode bridges in total. 2xLM317 for 5V and 2xLM723 for 15v (with 50mA foldback). All 4 sources are "galvanic" isolated, thus I have 4 voltages with 4 grounds. You may interconnect those sources as you wish (imagine 4 separate batteries with 15V, 15V, 5V, 5V). With a lab voltage source there is always the problem how to limit the current such you do not smoke your chips when experimenting. With bigger pcb I would use 4x 723, or similar, in order to provide a foldback.

Very nice, I like the flexibility a design like that offers.

[tunk]

I guess your 35VA 18V transformer will give you ~1+1A, not 2+2A.

[Vovk_Z]

Is there a difference in the ripple phase, between this and the more conventional single rectifier approach?

The ripple is the same with two diode bridges. There can't be any ripple phase difference because they both are full-wave bridge rectifiers.

[Vovk_Z]

The data-sheet said 0.1μF ceramic or 1uF tantalum on the input that's why I chose it.

You need those caps near the IC regulator if main rectifier caps are far from the IC regulator. 'Far' I mean more than 7-10 cm.

[Vovk_Z]

I guess your 35VA 18V transformer will give you ~1+1A, not 2+2A.

Yes. 1A AC, which is about 0.7 A DC.

[julian1]

Is there a difference in the ripple phase, between this and the more conventional single rectifier approach?

The ripple is the same with two diode bridges. There can't be any ripple phase difference because they both are full-wave bridge rectifiers.

Thanks. I was (incorrectly) thinking it was input freq, rather than 2x input freq, fully rectified.

[bdunham7]

I'm aiming for 2A output x 2. I understand I need to increase transformer voltage to 18V if I want regulated 15v out.
The data-sheet said 0.1μF ceramic or 1uF tantalum on the input that's why I chose it. The 5600uF was close enough to 6000uF for rule of thumb of 2000uF per amp for ripple rejection.

EDIT: Transformer is 35VA 18V Talema 70074K I think is what I will use. I have not ordered parts or anything.

I'm glad I asked.  Your transformer is way too small (by a factor of 4 or 5) and you will need very large heatsinks on the regulators. 

[networkingdude]

I'm aiming for 2A output x 2. I understand I need to increase transformer voltage to 18V if I want regulated 15v out.
The data-sheet said 0.1μF ceramic or 1uF tantalum on the input that's why I chose it. The 5600uF was close enough to 6000uF for rule of thumb of 2000uF per amp for ripple rejection.

EDIT: Transformer is 35VA 18V Talema 70074K I think is what I will use. I have not ordered parts or anything.

I'm glad I asked.  Your transformer is way too small (by a factor of 4 or 5) and you will need very large heatsinks on the regulators.

Largest I can go is 50VA but you guys are right, my transformer is way too small.

[bdunham7]

Largest I can go is 50VA but you guys are right, my transformer is way too small.

Why do you need so much power for a headphone amp?  I would have thought 500mA was plenty.

And to answer your other question, the resistors are there to bleed off the capacitors when the power is off.  They don't have to be large, just use a resistance high enough that the power dissipation is low.

20V/2.2K = 9.1mA and 182mW.  So a 2.2K 1/4 watt would be fine if you wanted this feature--which isn't a bad idea.

[networkingdude]

Largest I can go is 50VA but you guys are right, my transformer is way too small.

Why do you need so much power for a headphone amp?  I would have thought 500mA was plenty.

And to answer your other question, the resistors are there to bleed off the capacitors when the power is off.  They don't have to be large, just use a resistance high enough that the power dissipation is low.

20V/2.2K = 9.1mA and 182mW.  So a 2.2K 1/4 watt would be fine if you wanted this feature--which isn't a bad idea.

I absolutely don't need that just for a headphone amp, my thoughts were to make it larger in case I want to connect something larger in the future.

[fcb]

I will remove D3 and D6, I can see how they are redundant.

Don't get rid of them. It's more reliable with them.

D3/D6 protect against reverse polarity. This is an unlikely or potentially impossible scenario (only the OP knows for sure what’s connected, but it’s described as an audio amplifier). In that case it would technically be less reliable if fitted.

As for the size of the input capacitors, this isn’t so straightforward.  With the circuit as is, the datasheet reckons you’ll achieve something like 86dB of rejection (quite a lot, but datasheets can fib), so any ripple you get will be attenuated by 86dB (this will drop hugely as the frequency goes up).  Then you’ve got the CMRR of the amplifier, which might be perhaps 40-70dB, which will mean that you’ll be reducing 50/60Hz ripple by perhaps 130+dB, far in excess of what you’ll hear or be able to measure.

So going mad with the input capacitors will likely not help.  Reducing the HF component of the noise (from the diode bridge, transformer noise, mains borne interference) will be as (or more) important - so adding an LC filter after the each rectifier might be a better use of space/BOM cost.  Look up the rejection graphs at higher frequencies.

As for the topology and ‘ground loops’ - nothing wrong with the design, you effectively have two separate floating supplies, commoned at a single point.  In some senses it’s a better design as you’ll get the same performance out of each (not that a LM317/LM337 complementary supply is has wildly different +ve / -ve performance).

[Kibabalu]

I have to repeat it: You bypassed the diodes D1 and D4 and the voltage controllers IC1 and IC2 by a shortcut! Examine your schematic exactly!

And follow the link in #3!

[bdunham7]

I have to repeat it: You bypassed the diodes D1 and D4 and the voltage controllers IC1 and IC2 by a shortcut! Examine your schematic exactly!

And follow the link in #3!

Could you draw the short on the schematic and post it?  I don't see any short across those components.

[tunk]

Yes. 1A AC, which is about 0.7 A DC.

Ignorant question: is this a rule of thumb, or is it a calculation?

[networkingdude]

I have to repeat it: You bypassed the diodes D1 and D4 and the voltage controllers IC1 and IC2 by a shortcut! Examine your schematic exactly!

And follow the link in #3!

I can not see where there is a problem. Most of this schematic is directly from data-sheet. I will check the link again.

EDIT: Are you aware that this is a dual secondary transformer and these power supplies are both positive voltage regulators only as they are fully isolated from one another? I get what your saying is true when using a center tapped transformer with a negative regulator but this is not the case here.

[Kibabalu]

jahimmelhergottsakra

(a bavarian curse ;-)

[Kibabalu]

And follow link in post #3!

You'll find a design suitable for the combination LM350/LM333.

Or use the design attached to this post (excuse the poor quality of the picture). It works well and outputs a nice smooth output voltage, even if you feed it from a switch-mode power supply. If you need 3A, then replace LM317/LM337 with LM350/LM333.

[networkingdude]

And follow link in post #3!

You'll find a design suitable for the combination LM350/LM333.

Or use the design attached to this post (excuse the poor quality of the picture). It works well and outputs a nice smooth output voltage, even if you feed it from a switch-mode power supply. If you need 3A, then replace LM317/LM337 with LM350/LM333.

I must reiterate that I do not wish to use a negative regulator in this design.